Predictive modelling and identification of key risk factors for stroke using machine learning.

Area of Science:

• Cardiovascular Medicine
• Biostatistics
• Machine Learning in Healthcare

Background:

• Stroke is a primary cause of global mortality, necessitating improved early detection and prevention.
• Accurate stroke prediction is hindered by imbalanced and missing data, complicating risk factor identification.

Purpose of the Study:

• To develop and evaluate advanced machine learning models for stroke risk prediction.
• To address challenges posed by imbalanced and missing data in stroke prediction models.

Main Methods:

• Imputation techniques were used for missing data, and Synthetic Minority Oversampling Technique (SMOTE) for data imbalance.
• A range of advanced models were evaluated using k-fold cross-validation on diverse datasets.
• A Dense Stacking Ensemble (DSE) model was developed, utilizing fine-tuned advanced models.

Main Results:

• Key predictors for stroke include age, BMI, glucose levels, heart disease, hypertension, and marital status.
• The DSE model achieved over 96% accuracy on various datasets.
• AUC scores reached 83.94% on imbalanced imputed data and 98.92% on balanced data.

Conclusions:

• The Dense Stacking Ensemble (DSE) model demonstrates superior performance for stroke prediction compared to previous research.
• The DSE model shows significant potential for enhancing early stroke detection and improving patient outcomes.